coincided with a rapid rise in criminal activity and incarceration rates for them, especially relating to the drug trade (Freeman, 1994; Piehl, 1995).
Overall, this pattern of falling employment and wages for the less educated and rising employment and wages for the more educated strongly suggests that employers' demands have shifted toward workers with higher education more rapidly than the relative supplies of the two groups have been able to adjust. Indeed, relative increases in the supply of college graduates actually slowed during the late 1970s and early 1980s, reflecting an earlier drop in college enrollment rates as well as the declining number of young people in the ''baby bust" cohort. These enrollment drops appear to have contributed to the market premiums that college graduates have enjoyed (Katz and Murphy, 1992). More recent evidence suggests that enrollment rates have recovered somewhat since then (in response to the rising returns for education) and are likely to continue to rise (Mincer, 1994). But few observers expect these increases to be sufficient to fully reverse the recent increases in earnings gaps between education groups, especially since the relative demand for college-educated workers will continue to grow (Bishop, 1995).
Why does the demand for workers with higher education continue to rise? Growing international competition and the exodus of manufacturing plants from the United States to lower-wage areas are frequently mentioned as major reasons for these changes. But these labor market changes have occurred in nonmanufacturing (or nontraded goods and services) as well as manufacturing industries; even among the latter, most of the changes have occurred within particularized manufacturing industries rather than between industries whose relative employment levels are affected by trade (Berman et al., 1994; Freeman, 1995a).
The most likely suspect for this shift in employer demand is a rapid increase in the pace at which employers have implemented various technological changes. Computer-related technologies (such as CAD-CAM and robotics) have clearly transformed many manufacturing processes. At the individual level, computer use is clearly correlated with higher educational attainment or wage levels among workers (Krueger, 1993), and at the industry level, industries in which technological changes have been most rapid have also experienced the largest declines in employment among less educated production workers (Berman et al., 1994).
Of course, technological change need not always be "biased" toward more highly educated workers (Goldin and Katz, 1995), and examples can certainly be found of computers leading to lower rather than higher skill requirements for workers. But, on average, rising capital intensity and technological improvement seem more likely to be associated with higher demand for educated labor in the United States (Bartel and Lichtenberg, 1987; Hamermesh, 1993). Changes in the organization of production that at least some firms have chosen to undertake, such as total quality management and other "high-performance" workplace activities, might also increase employers' demand for education (e.g., Ichniowski et
al., 1995; Cappelli, 1995b, 1996). A variety of industry case studies seem to corroborate that this occurs with technological change and/or workplace reorganization (e.g., Bureau of Labor Statistics, 1984, 1986; Bailey, 1990; Levy and Murnane, 1995).
Thus, the data strongly suggest that employer demand for more educated workers has risen relative to their supplies in recent years. The case for a skills gap or mismatch in education levels can therefore be made. This is true even though overall unemployment levels in the United States are not high, especially relative to European countries. Apparently, the relative demand shifts in this country have largely resulted in low wages for the less educated (except among high school dropouts and young blacks, as noted above), while in Europe they have mostly resulted in lower employment rates (e.g., Freeman and Katz, 1994).12
Finally, there appears to have been no strong trend toward greater or lesser overall job stability in recent years, as measured by turnover rates or job tenure (i.e., time with a single employer). However, average tenure has declined somewhat among the lowest-wage employees. Whatever the causes of these trends, they certainly do not imply a greater willingness among employers to invest in job training for their low-tenure employees.13
A variety of caveats might be mentioned here. For one thing, although not all young college graduates will easily find employment, especially at relatively high wages (Hecker, 1992), they are, on average, a good deal more likely to do so than are those without college degrees, and the gaps between the abilities of these two groups to do so have certainly risen in the past 10 to 15 years (Bishop, 1995; Murnane, 1995). Furthermore, the rising relative demand for college-educated workers is not the only reason for the declining wages of the less educated. Falling rates of union membership, falling real minimum wage levels, and rising immigration all appear to have contributed to this development (e.g., Freeman, 1995c; Dinardo et al., 1995; Jaeger, 1995). Furthermore, the rise in labor market inequality has occurred even within education groups as well as within every other observable demographic category, and these increases remain not well understood (Levy and Murnane, 1992).
Finally, the falling wages of less educated workers in the United States reflect
the recent decline in overall earnings and productivity growth as well as a widening of wage inequality—that is, a stagnating mean (or median) level of earnings along with growing variance around that mean. The explanations for this overall stagnation are not widely understood, nor is there much consensus regarding future trends in this area. (For a very readable summary of trends and the historical perspective on this question, see Madrick, 1995.)
Despite these questions and caveats, the evidence strongly suggests that a rise in the demand for educated workers relative to their supply has widened the gaps in relative earnings and employment rates between more and less educated workers, consistent with a "skills gap."
Test Scores
The fact that earnings inequality has risen even within educational categories raises the question of whether employer demand for skills measured along other dimensions has risen as well (e.g., Juhn et al., 1993).
One measure of individual abilities that varies within educational categories can be found in scores on tests of cognitive ability. A long and sometimes controversial literature has appeared over several decades on the extent to which test scores actually measure cognitive abilities, on their environmental versus hereditary determinants, and on their correlations with earnings (e.g., Jencks, 1972; Herrnstein and Murray, 1994; Goldberger and Manski, 1995; Hauser and Carter, 1995). But some important new evidence on the labor market effects of test scores has emerged in the past few years. For one thing, Murnane et al. (1995) have found that the returns to test scores (especially for math) rose significantly during the 1980s. Indeed, they found that the entire increase in returns to college education for females can be accounted for by the increased return to test scores during that period. Murnane et al. also found that wage returns to test scores rose with the number of years of labor market experience. This reflects the fact that test scores (and cognitive abilities more generally) are not observable by employers at the time of hiring. But the rising returns to these scores with experience indicate that they may well be correlated with subsequent labor market performance, which employers can more easily observe over time.14
Other studies (e.g., O'Neill, 1990; Ferguson, 1993; Neal and Johnson, 1995) have also found that differences in hourly earnings between whites and blacks (after educational differences have been controlled for) can be accounted for largely by differences in the scores of the two groups on armed forces qualifying tests. Effects of test scores on the relative employment rates of whites and blacks have also been found (e.g., Rivera-Batiz, 1992), though a substantial racial difference